Monitoring Hybrid/Private Cloud with Prometheus and Grafana

This tutorial will guide you through the process of setting up Prometheus and Grafana to monitor Qdrant databases running in a Kubernetes cluster used for Hybrid or Private Cloud.

Prerequisites

This tutorial assumes that you already have a Kubernetes cluster running and a Qdrant database deployed in it, using either a Hybrid Cloud or Private Cloud deployment. You should also have kubectl and helm configured to interact with your cluster.

Step 1: Install Prometheus and Grafana

If you haven’t installed Prometheus and Grafana yet, you can use the kube-prometheus-stack Helm chart to deploy them in your Kubernetes cluster.

A minimal example of installing the stack:

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts

helm install prometheus prometheus-community/kube-prometheus-stack --namespace monitoring --create-namespace

This command will install Prometheus, Grafana, and all necessary components into a new monitoring namespace.

Step 2: Configure Prometheus to Scrape Qdrant Metrics

To monitor Qdrant, you need to configure Prometheus to scrape metrics from the Qdrant database(s). You can do this by creating a ServiceMonitor resource in the host Kubernetes cluster.

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: qdrant-cluster-exporter
  namespace: qdrant
  labels:
    release: prometheus
spec:
  endpoints:
  - honorLabels: true
    interval: 60s
    port: metrics
    scheme: http
    scrapeTimeout: 55s
  jobLabel: app.kubernetes.io/name
  namespaceSelector:
    matchNames:
    - qdrant
  selector:
    matchLabels:
      app.kubernetes.io/instance: qdrant-cluster-exporter
      app.kubernetes.io/name: qdrant-cluster-exporter
---
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: qdrant-operator
  namespace: qdrant
  labels:
    release: prometheus
spec:
  endpoints:
  - honorLabels: true
    interval: 60s
    port: metrics
    scheme: http
    scrapeTimeout: 55s
  jobLabel: app.kubernetes.io/name
  namespaceSelector:
    matchNames:
    - qdrant
  selector:
    matchLabels:
      app.kubernetes.io/name: operator

This example assumes that your Qdrant database and the cloud platform exporter are deployed in the qdrant namespace. Adjust the namespaceSelector and namespace fields according to your deployment.

The release: prometheus label must match the name of your Helm release. By default, kube-prometheus-stack only picks up ServiceMonitor resources labeled with its own release name. If you installed the chart under a different name, change the label to match.

Apply both resources:

kubectl apply -f servicemonitors.yaml

These two monitors cover the cloud platform exporter and the Operator. To scrape other components, see Networking, Logging & Monitoring.

Step 3: Check That Prometheus Sees the Targets

Before you build any dashboards, confirm that Prometheus discovered the Qdrant targets and can query a metric from them.

Port-forward the Prometheus service:

kubectl --namespace monitoring port-forward svc/prometheus-kube-prometheus-prometheus 9090

Open http://localhost:9090/targets in your browser. You should see two serviceMonitor/qdrant/... entries, one for the cluster exporter and one for the Operator, and every endpoint under them should have the state UP.

If you prefer the command line, curl -s http://localhost:9090/api/v1/targets returns the same information as JSON. Look for health: "up" in each entry of data.activeTargets.

Then query a Qdrant metric. The Operator exposes the status of every cluster it manages:

curl -s 'http://localhost:9090/api/v1/query?query=qdrant_operator_cluster_phase' | jq '.data.result'

A non-empty result means metrics are flowing. If a target is missing, check that the release label matches your Helm release and that the namespaceSelector and selector labels match your Services. If a target is listed but down, check that the metrics port name exists on the Service and that network policies allow Prometheus to reach the Pod.

Step 4: Access Grafana

Once Prometheus is configured to scrape metrics from Qdrant, you can access Grafana to visualize the metrics.

Get the Grafana ‘admin’ user password by running:

kubectl --namespace monitoring get secrets prometheus-grafana -o jsonpath="{.data.admin-password}" | base64 -d ; echo

Access the Grafana dashboard by port-forwarding:

export POD_NAME=$(kubectl --namespace monitoring get pod -l "app.kubernetes.io/name=grafana,app.kubernetes.io/instance=prometheus" -oname)
kubectl --namespace monitoring port-forward $POD_NAME 3000

Now you can open your web browser and go to http://localhost:3000. Log in with the username admin and the password you retrieved earlier.

Step 5: Import the Qdrant Dashboards

Qdrant publishes two example Grafana dashboards in the qdrant-cloud-grafana-dashboard repository. For Hybrid Cloud, import both:

For Private Cloud, use the Operator dashboard.

The dashboards are not versioned per Operator release. One Operator file covers Operator v2, and its queries expect Pods named qdrant-operator-v2-*, so they do not match Operator v1. The version field inside a dashboard JSON is Grafana’s own revision counter. Always import the current files from the repository.

To import a dashboard:

1. In Grafana, go to “Dashboards”, then click “New” and “Import”.
2. Upload the dashboard JSON file from the repository, or paste its contents.
3. Select the Prometheus data source that scrapes your Qdrant namespace, then click “Import”.

The dashboards need the cloud-specific metrics scraped in Step 2. They do not work with metrics from the open-source Qdrant container image.

Troubleshooting: Panels Show No Data

A blank panel usually means that the series of its query are not in the Prometheus data source you selected. Work through these checks:

1. Check the data source. Select the Prometheus that scrapes the Hybrid Cloud namespace when you import. Copies of the dashboards from before October 2026 have internal data source IDs hardcoded, which leaves every panel empty in another Grafana. Import the current files.
2. Check the cluster dropdown. The database dashboard lists clusters from qdrant_operator_cluster_info_total. If the Operator target on port 9290 is down, every panel on that dashboard is blank. Only Operator v2 exposes this metric.
3. Check the targets. Every Qdrant target must be UP, as in Step 3. A ServiceMonitor whose release label does not match your Prometheus Helm release never becomes a target.
4. Check the names and the network. The queries expect Operator Pods named qdrant-operator-v2-* and database Pods named qdrant-<cluster-id>-*. A different Helm release name, or a network policy that blocks Prometheus from the Operator metrics port, gives the same empty result.
5. Find the missing metric. Open the empty panel, choose “Edit”, copy the metric name from the query, and run it in Prometheus. For example:

qdrant_operator_cluster_info_total
container_memory_working_set_bytes{namespace="qdrant"}
kubelet_volume_stats_used_bytes{namespace="qdrant"}
kube_pod_container_info

Missing one source blanks only the panels that use it:

If you use Chronosphere, panels that use Grafana’s classic transformations, such as JoinByField, stay empty even when the metrics exist. See the open issue in the dashboard repository.